Human Activity Recognition with Trajectory Data in Multi-floor Indoor Environment

In pervasive and context-awareness computing, transferring user movement to activity knowledge in indoor is an important yet challenging task, especially in multi-floor environments. In this paper, we propose a new semantic model describing trajectories in multi-floor environment, and then N-gram model is implemented for transferring trajectory to human activity knowledge. Our method successfully alleviates the common problem of indoor movement representation and activity recognition accuracy affected by wireless signal calibration. Experimental implementation and analysis on both real and synthetic dataset exhibit that our proposed method can effectively process with indoor movement, and it renders good performance in accuracy and robustness for activity recognition with less calibration effort.